Good day, everyone, and welcome to the Vicor earnings results for the fourth quarter and the year ended December 31st, 2018, which is hosted by James Simms and Dr. Patrizio Vinciarelli. My name is Shelley, and I'm your event manager. During the presentation, your lines will remain on listen only, and if you require assistance at any time, just key star zero on your telephone and a coordinator will be happy to assist you. I would like to advise you all the conference is being recorded for training purposes and replay purposes. Now I'd like to hand the call over to James Simms. Please go ahead.
Thank you, Shelley. Good afternoon, everyone, and welcome to Vicor Corporation's earnings call for the fourth quarter and year ended December 31st, 2018. I'm Jamie Simms, CFO, and here with me in Andover are Patrizio Vinciarelli, Chief Executive Officer, and Dick Nagel, our Chief Accounting Officer. After the markets closed today, we issued a press release summarizing our financial results for the three and 12-month periods ended December 31st. This press release has been posted on the investor relations page of our website, www.vicorpower.com. We also filed a Form 8-K today related to the issuance of the press release. As always, I remind listeners this conference call is being recorded and is the copyrighted property of Vicor Corporation. I also remind you various remarks we make during this call may constitute forward-looking statements for purposes of the safe harbor provisions under the Private Securities Litigation Reform Act of 1995.
Except for historical information contained in this call, the matters discussed on this call, including any statements regarding current and planned products, current and potential customers, potential market opportunities, expected events and announcements, construction plans, as well as forecast sales growth, spending, and profitability are forward-looking statements involving risks and uncertainties. In light of these risks and uncertainties, we can offer no assurance that any forward-looking statement will in fact prove to be correct. Actual results may differ materially from those explicitly set forth in or implied by any of our remarks today. The risks and uncertainties we face are discussed in Item 1A of our 2017 Form 10-K, which was filed with the SEC on March 9th, 2018. We will file as required our 2018 Form 10-K by this Friday, March 1st, 2019.
Please note the information provided during this conference call is accurate only as of today, Tuesday, February 26th, 2018. Excuse me, '19. Vicor undertakes no obligation to update any statements, including forward-looking statements made during this call, and you should not rely upon such statements after the conclusion of the call. A replay of the call will be available beginning at midnight tonight through March 13th. The replay dial-in number is 888-286-8010, followed by the passcode 10247812. In addition, a webcast replay of today's call will be available shortly on the investor relations page of our website. I will start this afternoon's discussion with a review of our financial performance for the fourth quarter and the full year. Dick will comment on tax matters, and Patrizio will follow with his remarks, after which we will take your questions.
Beginning with consolidated results, as stated in today's press release, Vicor recorded total revenue for the fourth quarter of $73.7 million, representing a sequential quarterly decline of 5.5% from the $78 million recorded for the third quarter, and an increase of 25.4% over revenue recorded for the fourth quarter of 2017. On a full-year basis, total revenue for the year was $291.2 million, 27.8% higher than the total revenue for 2017. As other companies have reported, December quarter results fell below expectations due to several factors, most notably weaker demand in China and softness in data center spending. For Vicor, Q4 bookings showed evidence of weakening demand. Fourth quarter revenues were influenced by upwards of a $5 million value of shipments rescheduled into 2019. In a moment, Patrizio will address booking trends and specific circumstances influencing our outlook.
Last quarter, we began providing a breakdown of revenue by advanced and legacy product categories. The advanced products category represents the sum of the products of our Picor and VI Chip operating segments, reflecting proprietary patent advances in factorized power distribution, power conversion topologies, control systems, power semiconductors, and advanced power packaging. Legacy products are those associated with our well-established brick business unit operating segment, serving a broad range of classic distributed power applications with integrated converter modules, configurable standard product assemblies, and custom solutions. Going forward, we'll be referring to these products simply as brick products in our remarks and in our filings. We will continue to report our operating segments in our 10-K and 10-Qs as we have for some time as the BBU, VI Chip, and Picor operating segments, reflecting inter-segment transactions, our financial reporting structure and management oversight.
We believe the advanced product and brick product designations, net of intercompany inter-segment transactions, are appropriate for summarizing our progress in implementing our strategic transition from a relatively high-mix, low-volume model, serving a highly fragmented and diverse customer base, largely in mature market segments, to a relatively low-mix, high-volume model, serving OEMs and their manufacturing partners, focusing on high-growth opportunities with a portfolio of highly differentiated products. For the fourth quarter, advanced product revenue was flat sequentially, rising less than 1% over the third quarter, despite the program delivery rescheduling previously referenced. Advanced product revenue for the fourth quarter of 2018 rose 21.1% from the fourth quarter of 2017. As a percentage of total quarterly revenue, advanced products contributed 36.7%, up from 34.6% of total revenue for the third quarter.
For full year 2018, advanced products revenue totaled $104.6 million, a 36.8% increase over the total for 2017, and represented 35.9% of 2018 total revenue versus 33.6% for 2017. Fourth quarter revenue for brick products declined 5.5% sequentially, but rose 28.1% over the fourth quarter of 2017. As a percentage of total quarterly revenue, brick product sales represented 63.3%, in contrast to 65.4% for the preceding quarter. For full year 2018, brick product revenue totaled $186.6 million, a 23.3% increase over the total for 2017 and represented 64.1% of 2018 total revenue versus 66.4% for 2017. Quarterly international revenue declined 8.1% sequentially and represented 60.5% of total revenue. Advanced products exports rose 6.5% sequentially and represented 46% of exports for the fourth quarter, indications of the stability of our shipments for 48-volt to point-of-load programs with Asian ODMs and CMs.
Brick products declined 17.8% sequentially and represented 53.9% of exports for the fourth quarter, down sequentially from 60.3%, reflecting bookings declines from our Chinese and European distributors, which largely serve industrial customers. Geographically, total revenue for 2018 breaks down as follows: North America represented approximately 38%, China, including Hong Kong, represented approximately 37%, Asia Pacific, excluding China, represented approximately 13%, Europe and the Middle East represented 9.5%, and all other geographies sum to 2% of total revenue. For revenue across 2018, we recorded increases across all geographies, reflecting higher unit volumes and improved pricing. Exports to China of both advanced products and brick products accelerated early in 2018 and finished the year-over-year growth of approximately 34%. North American revenue increased approximately 32%, driven by strength in aerospace and defense electronics, balanced between advanced products and brick products.
Exports to Asia Pacific, including China, grew approximately 16%, and exports to Europe increased by 15%. For the fourth quarter and the full year, we had one 10% customer, New Power Electronics, our largest authorized distributor in China. For the fourth quarter and the full year, our five largest customers, including New Power and two other distributors, represented 31.3% and 36%, respectively, of total revenue. Recall that we report export volumes on the basis of ship-to address. As such, the 10% customer concept may not provide a complete view of customer concentration. Distributors obviously have many downstream customers with diversified applications. In contrast, an OEM using a relatively small number of our SKUs usually spreads a program's volumes across multiple contract manufacturers and combined shipments for a given program for a given period may exceed 10% of revenue.
Turns volume, that is, orders recorded and shipped within the quarter, was approximately 20% of fourth quarter revenue, up from approximately 18% of third quarter revenue. Lower turns volumes throughout 2018 have been a reflection of extended lead times for certain raw materials, which have resulted in extended scheduled delivery times for customer orders. However, given improved availability of some raw materials, we have been able to accommodate some customer requests for accelerated delivery, which is a sign of improved supply chain visibility. As mentioned, Patrizio will address bookings in his remarks, but I want to emphasize the following, which I've spoken to before. Bookings and shipment patterns can differ, sometimes meaningfully, for advanced products and brick products. Advanced products, bookings, and shipments thus far have reflected the low mix, high volume requirements of a concentrated customer base with individual high-volume orders scheduled for delivery over multiple quarters.
Design wins for advanced products can require upwards of a year of engagement with an OEM, and production volume orders from one or more contract manufacturers may not be placed for a similar period. To illustrate, a hypothetical OEM program may be planned for a three-year life before a major redesign. Such a program typically involves multiple contract manufacturers, which place orders with us based on the OEM's production schedules. In any year of the program, we might receive relatively few orders, but the orders are for high volumes to be shipped over several quarters. Such shipments may be subject to rescheduling by the OEM based on revisions of its demand forecast or by the contract manufacturer based on their factory loading and supply chain visibility.
As we seek to expand the number of OEMs and their programs with which we're involved, we expect to smooth out this lumpiness of the cycle for advanced products. Until then, listeners should consider the relatively large impact of a program change, notably shipment rescheduling, on our results as we experienced in the fourth quarter. In contrast, brick products generally are high mix, low volume, serving a statistical customer base of nearly 10,000 customers. Competition for design wins is less intense and generally of short length for brick products, given our mass customization manufacturing model, which enables us to meet customer requirements that catalog-based companies cannot. Many brick products have been designed into customer end products for a long time, and such customers typically are in relatively stable, mature markets with well-recognized demand patterns.
While a few large customers place program or blanket orders, brick product orders are generally smaller, averaging less than 100 units and scheduled over weeks and months, contributing to a smooth booking pattern across business cycles. There are anomalies to booking patterns as we believe we are currently experiencing. For example, in China, where the imposition in September of an additional 10% import tariff on U.S. manufactured goods made trading conditions challenging, the threat of another round of higher tariffs to be imposed if the current trade negotiations break down, caused customers to postpone orders in the fourth quarter and to date in the first quarter of 2019. A definitive positive resolution of the uncertainty about additional tariffs may result in a resumption of vigorous order flow. I'll now turn to product profitability.
Consolidated fourth quarter gross profit margin as a percentage of revenue was 45.9%, largely reflecting lower product volumes and inefficiencies caused by shifting product mix. Advanced product volumes were flat, with increased VI Chip shipments offset by lower Picor shipments. Changing mix was a contributing factor for both product categories. While our third quarter consolidated gross profit margin rose to 50%, stable mix was a strong contributor, as was relatively high volume. For the fourth quarter, manufacturing performed well under difficult conditions of changing demand, and we remain confident of the scalability of our business model. Also during the quarter, we were successful in meeting our needs for raw material inventories as vendor lead times shortened. Section 301 tariffs on our Chinese imports did not have a material impact on our procurement costs during the quarter.
The cost going forward may not be inconsequential given the volume of components we source from China. When possible, we are seeking alternate non-Chinese vendors. As indicated last quarter, we filed requests with the U.S. government for exclusions from tariffs on a limited number of components for which no alternate vendor exists. We have not received any response yet. We continue to monitor the impact of tariffs on the cost of imports from China. If this impact becomes meaningful, we may add a tariff surcharge to the selling price of our products. Turning to operating expenses, fourth quarter operating expenses rose 3.2% sequentially. For 2018, total OpEx rose 3.4% over 2017, in line with our expectations, with customary increases in compensation and outlier increases in audit and legal expenses offset by a decline in R&D prototyping costs.
On a relative basis, Q4 operating expenses reversed a seven-quarter trend, actually rising as a percentage of quarterly revenue to 36.3% from 33.3% the prior quarter. This is due in part to the unanticipated decline in revenue. R&D expenses rose 3.5% sequentially and from 13.7% to 15% of quarterly revenue, with the only noteworthy spending variance being another decline in prototyping costs. Marketing and sales expenses rose 3.1% sequentially and from 13.6% to 14.9% of quarterly revenue, with no noteworthy spending variance. G&A expenses rose 2.6% sequentially and from 6% to 6.5% of revenue, also with no noteworthy spending variance. Operating income declined sequentially to 9.5% of revenue from 16.7% for the prior quarter. For the year 2018, operating income was 11.1% of revenue, up from a negligible operating loss for 2017. On an absolute basis, these results are in line with our recent statements regarding spending trends.
We continue to expect operating expenses to decline as a percentage of revenue, while expanding at low single-digit percentage rates in absolute terms, largely driven by compensation costs. As stated before, our long-term model is to reduce total operating expenses to 30% of revenue. I will now turn to Dick for a quick overview of our tax position. Dick?
For the fourth quarter, our effective tax rate was 5%. We recorded a net provision of $363,000. For the full year, our effective tax rate was 3.3%, reflecting a cumulative provision of just under $1.1 million, the majority of which consisted of state and foreign tax expense. During prior conference calls, we have explained our perspective on the valuation allowance we maintain against the value of our domestic deferred tax assets under the assumption sequential quarters of improved profitability would allow us to conclude the release of the allowance would be appropriate. While our outlook remains positive for sustained profitability, the results of the third and fourth quarter, combined with generally poor near-term visibility, led us to conclude a release of the valuation allowance, which stood at approximately $30 million as of December 31, 2018, would be premature.
Pursuant to the requirements of ASC 740, we have left the full valuation allowance in place. We will continue to assess the potential release of the allowance later in 2019. The composition of our deferred tax assets shifted further at year-end as the company has been utilizing available net operating loss carryforwards and tax credits to offset taxes due on taxable income throughout the year and consumed a major portion of its federal NOL balance in 2018, leaving federal and state R&D tax credits, along with other tax credits, reserves, and other accounts as the balance of our DTAs. As I previously mentioned, we believe it is more likely than not we will release some portion, if not all, of the valuation allowance later in 2019. Jamie?
Thank you, Dick. Returning to the P&L, we recorded fourth quarter net income after minority interest of $6.9 million, down from the $13 million recorded for the prior quarter and up from the $1.6 million recorded for the fourth quarter of the prior year. For 2018, net income, again, after minority interest, totaled $31.7 million, up from an essentially breakeven level for 2017. Diluted EPS for the fourth quarter totaled $0.17, down from the third quarter's $0.32 and up from the $0.04 per share recorded for the fourth quarter of the prior year. Full-year diluted EPS was $0.78, in contrast to last year's breakeven figure. Our diluted share counts for EPS calculations were 40,729,000 for the full-year EPS and 40,981,000 for the fourth quarter EPS. Turning to the balance sheet, cash and cash equivalents sequentially increased to $2.4 million for the fourth quarter, ending at $70.6 million.
On a year-over-year basis, cash and cash equivalents increased by $26.3 million, net of significant investments in equipment to expand manufacturing capacity made late in the year. The fourth quarter cash increase reflects operating cash flow of $13.3 million for the quarter, a decline from the $14.3 million recorded for the third quarter. A positive working capital swing offset the net income decline for the fourth quarter. Capital expenditures rose to $11.3 million from the prior quarter's $3.3 million, reflecting the substantial amount of new production equipment installed, as well as certain facilities upgrades. Capital expenditures for 2018 totaled $18.2 million, up from the prior year's $12.5 million. I'll return to capital spending and capacity in a moment. Trade receivables, net of reserves, totaled $42.8 million, down sequentially 2.6%. DSOs rose to 44 days from 41 days. At year-end, the trade receivables portfolio was, and remains today, sound.
Inventories, net of reserves, increased 9% sequentially to $47.4 million, largely reflecting rising material and component purchases to ship our advanced products backlog. Annualized inventory turns were stable at 3.6 times. Concluding my review of the fourth quarter, total employee headcount as of December 31st declined to 1,007 from 1,018 for the prior quarter, but rose from the December 31st, 2017 total headcount of 980. Full-time headcount was 976 at year-end, up two from the prior quarter and up six year-over-year. I'll now provide an update on our capacity expansion and planned capital expenditures. We have approximately $7 million of production equipment scheduled to be placed in service in the first half of 2019, which is expected to complete the first phase of capacity expansion that began last year. Completion of this investment will add approximately 35% to our chip manufacturing capabilities.
We are into the permitting phase of an 85,000 sq ft addition to our Andover manufacturing facility. We plan to break ground on this addition to our existing plant in 2019 and take occupancy in 2020. The planned addition of multiple manufacturing lines in this additional space, sequenced across 2020 and 2021, is planned to increase our advanced products capacity by an additional 100%. We are budgeting roughly $17 million for the construction, which will be paid through 2019 and conclude with occupancy. We are also budgeting an additional $12 million for two phases of production equipment installation, the first beginning at occupancy. The second phase of capital equipment spending will be scheduled six months ahead of initial demand for that capacity. Based on these budgets and our outlook for cash generation, we anticipate internally funding both the building and the two phases of investment in automated manufacturing equipment.
This sequenced approach to capacity expansion is designed to allow us to meet anticipated production scenarios by increasing our total Andover capacity to the $750 million revenue level by the end of 2020, while deferring, to the extent possible, incremental capital commitments and equipment depreciation charges. We are also exploring, with potential licensees, establishing an alternate source for advanced products through a separate manufacturing facility outside of the U.S. The licensee's facility would complement Vicor's initiative to address global demand for 48-volt power system solutions in both automotive and artificial intelligence applications. Turning to our outlook, Q1 revenue will be lower than Q4, while bookings are expected to increase sequentially with a Q1 book-to-bill ratio slightly above one. Bookings are forecast to increase at an accelerating pace as the year progresses, with revenues lagging one to two quarters. We expect to remain profitable and generate operating cash flow.
Having offered this limited guidance, I will remind listeners, as I do each time I speak with you, our operating and financial forecasts are subject to unanticipated changes, many of which are caused by factors and influences outside of our control. With that, I'll turn the call over to Patrizio.
Thank you. Fourth quarter bookings for all products declined 34% sequentially. Bookings for brick products declined approximately 15% sequentially, with particular softness in Chinese industrial markets. Geopolitical uncertainty contributed to reduced expectations from Europe. Until trade matters are resolved, our outlook for both Chinese and European industrial markets is cautious. However, confidence in U.S. markets remains relatively firm for both brick products and advanced products. Overall, bookings for advanced products declined sequentially as orders for our Power-on-Package solution for high current AI applications did not follow through from the progression of prior quarters. Near-term volume expectations for our first GPU Power-on-Package, or PoP, solution were reset by well-publicized forecast revisions from the lead customer for that solution.
Bookings for other 48-volt direct to point-of-load programs were also slower than anticipated in Q4, reflecting temporary softness in data center spending, with new programs scheduled to ramp in the second half of 2019. We remain confident of the opportunity before us in the rapidly developing artificial intelligence, supercomputer, and data center segments. A feverish pace of new program activity is setting the stage for substantial growth in bookings, revenues, and profitability. For Power-on-Package, we have secured major design wins in the AI space and in supercomputing, with production ramps expected in 2020. Our engagement with market participants continues to expand as companies developing advanced GPUs, TPUs and application-specific integrated circuits, specifically artificial intelligence ASICs, face fundamental power delivery challenges. With power delivery standing in the way of performance, Vicor's Power-on-Package has quickly established itself as the power system solution needed to overcome power conversion and power delivery challenges.
Our technological leadership is expanding as competition remains anchored by legacy 12-volt infrastructure, or variants on the theme of a so-called intermediate bus architecture, known to be fundamentally handicapped after repeated failed attempts at competing with 48-volt Factorized Power. Recognition of unrivaled technical leadership and enabling product capabilities are motivating broader adoption of Vicor's Power-on-Package solutions by leading developers of GPUs, TPUs, and every other high current ASIC. Anticipating rising current requirements from hundreds of amperes to thousands of amperes to fuel smarter AI processors, Vicor has developed Power-on-Package solutions using proprietary vertical power delivery, or VPD technology. AI design wins for high-performance VPD solutions evidence Vicor's ability through fundamental innovation to keep expanding the company's product leadership, causing competitors to have to play catch up with inadequate products that take up more than twice as much space in applications where there's no extra space to be had.
In space-constrained applications, Power-on-Package design wins set the stage for complementary front-end design wins, with liquid-cooled RFMs providing a nearly 10-fold increase in power density relative to conventional AC to 48 DC front-ends, including front-ends using GaN FETs. Even though GaN FETs are capable of switching at high frequency, in high voltage power conversion applications, the switching frequency of GaN FETs is limited by losses in power conversion circuitry to less than one-tenth the operating frequency of the RFM. As such, because of a speed limit imposed on high voltage GaN FETs by power conversion circuitry, front-ends using GaN FETs take up nearly 10 times more space than the RFM, which is not constrained by similar speed limits. GaN FETs also fail to add value, either in terms of performance or cost, to low voltage applications, specifically with the emerging 48-volt standard in automotive applications.
In automotive, Vicor is gaining significant traction with OEMs and tier 1 suppliers challenged by electrification, the advent of the 48-volt bus, and autonomous driving. Our value proposition includes high power density and modular flexibility. As recently reported, deployment of the most advanced Level 5 autonomous driving system will ramp substantially starting in 2020. Based on this and the trajectory of other early engagements, we expect that our growth in the automotive segment will, before too long, be comparable to the growth we're now experiencing powering artificial intelligence and servers in data centers. Across key growth markets, we're addressing major opportunities and scaling up capacity to support forecast production requirements. Let's now turn to questions. Operator?
Shelley?
Thank you. If you do have any questions, please key star then one on your telephone. Please make sure your line is unmuted when doing so. If you need to withdraw your question, please key star two. All questions will be answered in the order received, and you will be advised when to ask your question. All other lines remain on listen only. It's star then one if you have any questions. Thank you. Your first question comes from the line of Quinn Bolton. Please go ahead.
Hi, Patrizio. Hi, Jamie. Jamie, just wanted to start, I guess, first on a couple of business questions. It sounds like you mentioned there were some rescheduling of $5 million of orders from 2018 to sometime in 2019 from, I think, your largest Power-on-Package customer. Have those reschedulings, or have you seen any additional reschedulings quarter to date here in March, or has that forecast now stabilized?
I'll let Patrizio take it.
Time will tell whether their forecast has stabilized, but I think our exposure from a revenue and backlog perspective is limited and fundamentally tied to the $5 million that is shipping later this quarter.
Okay. That will actually ship here in Q1?
Q1 and very beginning of Q2. Yep.
Okay, great. I know you've referenced a couple of times in your prepared comments that visibility is low, but you did say you thought bookings and revenue would grow into the second half. I'm trying to understand, are you sort of anticipating that the March quarter would be the bottom, or do you really just not yet have enough visibility to call the bottom of this cycle?
We expect to see nearly 15% growth in bookings in Q1, and progression with higher bookings levels in Q2, Q3, and ending the year in Q4 near the $100 million level.
Okay, revenue you said typically follows that booking trend by one to two quarters.
This will follow the bookings by nearly two quarters, more than one quarter, closer to two quarters.
Okay, great.
Quinn, I also want to take exception with the use of the word cycle, because what happened with the rescheduling was very specific to a circumstance, and in general, it's very difficult for anyone to separate the events that are largely categorized as macroeconomic, when in fact, it's our point of view, notably in China, that the slowdown is most fundamentally associated with customers waiting for more certainty about tariffs. Does that help?
It does. Just last sort of business question, you've mentioned now on a couple of conference calls that you source a fair number of components in China, and that you're trying to find sources outside of China. Just wondering if you could give us an update on that initiative.
Well, we are realigning our supply chain in view of a number of considerations, obviously, the dependency from Chinese suppliers being one of them. It's not the only one. We believe that we are well-positioned going forward to ensure that we have continued supply, and importantly, continued supply at cost-competitive levels that support our expanding business model.
Great. The last one from me for you, Patrizio, just you mentioned the vertical power delivery or the geared MCMs give you the capability now to deliver more than 1,000 amps of current. Can you just sort of give us a quick overview? What is it about the geared MCMs or vertical power that allow you to get into the 1,500 amp or higher range with that technology? Thank you.
This technology essentially caps by nearly a factor of 10, the interconnect resistance between the Current Multipliers and the XPU being powered by the Current Multipliers. For instance, the well-advertised GPU application, the Current Multipliers are located on the sides of the GPU. There are tens of watts of power loss that take place in the current traveling sideways from the sides of the XPU to the bottom of the XPU. In vertical power delivery, the Current Multiplier or Current Multipliers, we have an application, for instance, with four Current Multipliers. The four Current Multipliers reside under the XPU, and that allows the current to travel a much shorter distance to feed the XPU in a much more efficient and uniform way. This may sound relatively obvious, but in reality, it's quite complex because there are obstacles standing in the way of providing vertical power delivery.
I'm not going to go into those technicalities, suffice it to say that it's not easy to do. It's been viewed as a complex task. We've overcome the fundamental challenges with proprietary developments that enable this capability. To be clear, lateral power delivery is not coming to a screeching halt. In fact, we have new applications with more advanced GPUs that will take it up to as much as 1,500, 2,000 amps, still with lateral power delivery. We also recently won, virtually without any effective competition, the first vertical power delivery opportunities were, again, the value proposition is one of much more efficient power delivery. One where, in effect, the level of AI capability is not gated by gating the current to the XPU.
Thank you.
Thank you. Your next question comes from the line of Matthew Vigneau. Please go ahead.
Hello. This is a similar question to Quinn's. In the past, management has mentioned that a robust economy wasn't needed to deliver growth. My question is, what really leads to confidence that this is mostly tariff-related and temporary versus the potential for a push-out of the data center upgrade cycles, and a push-out that could potentially last for a longer period of time? Thank you.
To be clear, the issue, or the headwinds as we refer to the circumstances of the fourth quarter, had to do with more than the situation with the tariffs on imports into China. They had to do, as mentioned in our prepared remarks, with one of our major applications for GPUs suffering setbacks that had nothing to do with China. More generally, with some level of softness in data center infrastructure spending. We believe, based on a variety of inputs we've received, that that softness is very temporary. These new programs that are starting to ramp in the middle of the year. Going back to your fundamental question of how do we ensure that we can deliver growth in good times, but also in not so good times, obviously not necessarily at the same level. The answer to that question is very simple.
By having kickass products that beat the competition dead, that's what we're doing in point-of-load applications. That's what we're doing in supercomputing applications. Before too long, that's what we're going to be doing in front-end applications. It's the reality that essentially, take as an example, the artificial intelligence chip market. That's a market, we sum , we value at around $1 billion. Everybody of significance in that market has come to Vicor for their solution. We're the only game in town. That's how we ensure long-term growth, and that's one market. There are other opportunities in other markets that we're also pursuing because we believe in a differentiated business model, one that is not predicated on one particular end market. We're building presence in automotive, we're building presence in other kinds of applications.
Before too long, as mentioned earlier, we're going to have very strong presence with front ends that are going to complement our superior point-of-load capability.
Do you have a follow-up, Matthew?
No, thank you so much.
Thank you.
Thank you. Your next question comes from the line of John Dillon. Please go ahead, John.
Hi, Patrizio. You mentioned the $100 million bookings level. I wasn't quite sure, are you talking about the second or third or fourth quarter for that level?
Fourth quarter.
Okay. It'll ramp from where we are. You expect 15% increase in bookings up in the first quarter, and it'll ramp up to the $100 million level, is that correct?
Yeah, that's the expectation at this point.
Okay. The 15% increase in bookings, I guess you're pretty far into the quarter, so I would imagine you have a pretty good feel for that. Is that pretty good?
Well, the quarter is not over and, as we've seen in the fourth quarter, surprises can happen and do happen. That's our forecast. I got confirmation this morning that that forecast is still good as of today.
Good. Okay. On your vertical power delivery products, this kind of brings me back to the IBM days. I believe on the start of the VI Chip, you provided the power supply right underneath the processor of the IBM computer. If I remember correctly, you actually enabled IBM to have a higher performance computer back then. Is that correct?
Yeah, that's really not vertical power delivery as enabled with the technology for which we recently got initial design wins. The problem we're solving with vertical power delivery is quite a bit more complex than in the early going of Factorized Power. It involves a changing of gears, so to speak, which is necessary in order to adapt from the cell pitch of our converter cells, which is around three millimeters or three and a half millimeters, to the pitch within an XPU or the power feed within an XPU, which in one instance is one millimeter pitch. In another instance, is down to 0.45 millimeter pitch. That change of pitch presents certain challenges.
Also, not to get too technical, just to give you a flavor of another challenge. The area under the XPU has historically been consumed by hundreds of bypass capacitors that are required in order to ensure that as the XPU undergoes rapid changes in current draw, which now are getting to literally 1,000 amp change on timescales of nanoseconds, that XPU voltage node is kept within very tight control range. The bank of capacitors has historically populated the area under the XPU and precluded power converters, Current Multipliers in particular, from occupying that same area. We also solved that problem. Fundamentally with the GCM, we're enabling a complete solution that delivers currents in potentially the thousands of amperes with very high bandwidth, in effect, occupying the same space that historically has been taken up by bypass capacitors.
You're obviously, like you mentioned before, you're saving a lot of power by doing this by a factor of 10, I think. There's resistance by a factor of 10.
Yeah. Interconnect resistance in a lot of power delivery application would typically be, let's say, 70 microohms. That may sound like a very small number, but when you're passing through that 1,000 amps, that's 70 watts worth of heat, which is a light bulb in and of itself. With vertical power delivery, that 70 microohms gets turned into eight microohms. You're essentially cutting that power loss by an order of magnitude. Therein lie 60, 70 watts of peak power dissipation savings that can be, in effect, utilized to not just provide more power to the XPU, but also to provide more control power to the XPU. Because as voltage nodes on 7 nanometer technology get to lower levels, we have applications now down to 0.4 volts. The processor is run nearly sub-threshold at 0.4 volts.
That's a very low voltage. Literally millivolts make a difference with respect to performance and operations. The headroom that is allowed in the power delivery path for drops associated with current delivery gets further and further reduced. That's why vertical power delivery is an enabling technology that ultimately results in higher performance AI processors. Needless to say, high performance is what this is all about, and that's why, you name it, every one of the major contenders for opportunity in that space comes to the place that has the technology.
That's incredible. You're saving 60 or 70 watts, you multiply that by the number of processors in a supercomputer or a data center, you're talking about some serious power savings.
Yeah. Now, to be clear, those are peak power draws. They're not necessarily the average power draws, the system has got to operate also under instantaneous peak power draws. These are complex issues, it's easy to trivialize them. I think it's fair to say that we've taken a very holistic view of the entire problem at the point-of-load and solved it. The solution comes back to 48 volt. Some people say 48 volt is going to take a long time to take root in the data center space. Absolutely not. 48 volt is being brought about not just because of the benefits of more efficient power delivery. I think ultimately the real driver is enabling the level of high performance that makes our customer solutions more competitive relative to their own competition. That's what will drive 48 volt adoption in data centers.
Ultimately, the whole market is also a very powerful driver, and that's a market with conversion to 48V is happening. There's no question about it.
Along the same lines of the vertical, you've said many times you free up pins, doesn't that also enable a higher performance computer because you have more I/O? Can't you place supporting chips closer to the processor so the time that electrons go from the processor to the memory or another processor are shorter? Doesn't that also make a faster, higher performance computer?
Yes, because generally speaking With lateral power delivery, you are confining the egress for the signal I/Os that, to your point, provide the needed bandwidth for communication with other devices. By going to the back, and obviously going there requires solving all the problems that stand in the way of vertical power delivery. By going to the back, essentially, the entire perimeter of the XPU is freed up. We have applications where these capabilities are going to be taken to new orders of magnitude of performance. They're literally going to blow away the level of performance of the best GPU of today.
Patrizio, that's the point that I'm excited about, because I don't think the average person, or I don't think the average engineer understands that a power supply can be an enabling technology. To me, that's what I get really excited about. I've taken up a lot of time, so I'm going to get back in the queue. Thank you very much.
Thank you. Any other question?
Yep. The next question is coming from the line of Alan Hicks. Please go ahead, sir. Your line is open now.
Yeah, good afternoon.
Good afternoon.
You've been talking about the opportunity in ASICs, the high amperage ASICs. What kind of opportunity do you have there, and what kind of applications are those?
We're not going to name customers.
As I suggested earlier, we have essentially every high current application that we know of, and that has generally been identified. I saw, for instance, the recent Barron's story on AI, projecting growth at the 50% per year level over the next 5 years, to, I think, a $34 billion market by 5 years from now. There's a lot of names in that story. We're doing business with virtually every one of them. That's because, again, in these applications, what's needed is a level of current delivery and performance that cannot be supported otherwise. What has happened over the last few years is something that I think many of us would have had a hard time predicting. It wasn't that long ago that an Intel processor could be fed with less than 100 amps, and could have provided all the performance that applications required.
Only in a matter of few years, we've gone from less than 100 amps to a few hundred amps, to 500, 600 amps, with peaks of 1,000 amps, to our highest current application as of a few weeks ago is going to be pretty close to 1,000 amps, and that's coming in 2020. This escalation is clearly driven by a need, and that is the need to, as suggested in the prepared remarks, fuel these devices that are running on finer and finer lithography with very, very large currents at lower and lower voltages. That is a compounded problem because, as suggested earlier, by going to a lower voltage, you run out of headroom for voltage error. You're enabling more current delivery with acceptable levels of power dissipation.
Obviously, if you cut the voltage enough, you can feed it with double the current for the same power consumption within the ASIC. Ultimately, all these applications are driven not by considerations of power savings. They're really driven by considerations of computing capability. Fundamentally, every competitor is looking to feed their ASIC with the most current that they can deliver and thermally managed. Thermal management is very capable, particularly with liquid cooling. The headroom that is made possible by advanced thermal management techniques has allowed this exponential rise in current consumption, which are ultimately a proxy for computing capacity.
Is the main opportunity in the AI area or supercomputing, or how does it compare to the GPU opportunity?
Well, it's in both. I think we are in all of the above. We're going to be in all of the above.
Okay. It's as big an opportunity as GPUs or CPUs?
Well, they all have a common denominator need, which we are, I believe, uniquely equipped to address. Their common denominator need is voltage delivery at less than one volt, typically 0.8, 0.7, 0.6, 0.5 volts. Our currents, as suggested earlier, are rising. They've risen by an order of magnitude in the last few years. For the time being, there's no end in sight as long as they can be thermally managed. With advanced thermal management techniques, you can easily get to applications that may consume, in one device, literally a few tens of thousands of amperes.
Okay.
There's a value proposition there. It's a supercomputer in a very small space that can do all kinds of wonderful things.
With the supercomputer market, in general, I know you get they're huge projects. Do you expect any wins this year?
We are engaged, again, with every significant competitor. They're all coming to Vicor for good reason, which is we can deliver a level of power density, current capability, power distribution efficiency, that nobody else can. This is enabling, as suggested in answer to an earlier question, it's enabling technology. We're not a supplier of commodity solution. We are a partner that enables a much higher level of performance that can be otherwise achieved.
Mm-hmm. I know you had the PEZY win a year or so ago. It was about an $8 million order, if I remember right. How big are these orders big for a big supercomputer?
There are many opportunities that are in the tens of millions of dollars per year range.
Mm-hmm. Okay. Thank you very much.
You're welcome.
The next question is coming from the line of James Liberman. Please go ahead, sir. Your line is open.
Thank you, and greetings. Could the adoption happen independent of the timing of autonomous vehicle rollout? In other words, would it just happen because you've got multiple supercomputers on board more and more cars? Is that going to be as equally a motivating factor?
We've been working with the company in that space, which we believe is the highest, the most capable autonomous driving system, a level 5 driving system. And scale output that should happen independently of other initiatives that we've been pursuing of late with the hiring of a VP for global automotive initiatives. But before too long, that opportunity is going to be complemented by a number of other opportunities that are not necessarily related to autonomous driving. They are related to the electrification of vehicles, the 48-volt system, IB systems that have dual 12 and 48-volt buses, a conversion between those buses, 48 to 12 to 48. There is a plethora of requirements and opportunities where very high density, lightweight, efficiency, modularity, scalability, cost-effectiveness, play a role with respect to creating opportunity.
Our team has been visiting OEMs in Japan, in Europe, in the U.S., and the universal response from all of these companies is they are blown away by the capability of our technology. They actually want us to be a tier 1 to them with respect to our capabilities. And we're pursuing a combination of initiatives to make that happen as quickly as possible. Word of caution here, nothing moves very quickly in automotive, and that's mitigated to some degree in today's automotive industry, which, as we know, is challenged to adapt to changing competitive landscape. It's not stronger moving at the glacial speed of earlier times, but it is still, for a variety of reasons, progressing in terms of the cycle of initial evaluation, designing, validation, and scale-up for production.
It's still moving at a slower pace than some of the market opportunities that we referenced earlier, which can turn from planting a seed to volume production in nine months or a year. I think the timescale that we should all be thinking of with respect to our presence in the automotive in a meaningful way, at the scale comparable to some of these other markets, that timescale is beyond the three, four years. It's not shorter than that.
Great. Thank you very much. That's quite helpful.
Thank you. If there's one more question?
Yes, there's a question left, and it is coming from the line of John Dillon. Please go ahead. Your line is open.
Hi, Patrizio. Can you give us a little more color on the front-end products? In particular, when do you think you'll start seeing a revenue pickup on those, and any kind of significant revenue from those products?
We're seeing more instances of combined opportunities with point-of-load. A recent example of a customer where we've been working for a very advanced point-of-load solution, wanting to incorporate an RFM front end. With the RFMs, as we currently have it, we have a solution that is based on earlier generation control. We are progressing rapidly to what we call 4G control. We're rolling out 4G in a variety of products, including a more advanced RFM that will be even denser and even more cost-effective. We see these front-end opportunities beginning to play out in scale over the next year. Initially with the existing RFM, which is based on older generation control. Once we introduce 4G-based solutions, we see, because of their cost attributes and even higher performance attributes, we see opportunities for bigger uptake.
When you say over the next year, for example, when do you expect to get maybe, let's say, $10 million in revenue? Would that be a year from today, or could that be six months from today, or?
I think it would be closer to a year from today.
In like six months, could you expect a couple million or $5 million, or?
Well, I think it's on the scale.
I'm just trying to get a.
It doesn't really move the needle. I think.
Gotcha
the way we should be thinking about this to avoid losing the forest for the trees, is what moves the needle in the short term is broad penetration of AI applications, supercomputing applications. That's moving the needle. I think front ends complementing those point-of-load applications will provide another wave of opportunities. As we discussed in the past, they can literally, at every one account, in every application, double the revenue opportunity. As I suggested earlier, the automotive initiative, that's even further into the future. In the short term, we're working our major opportunities, which are still in the center space. They're still with AI, GPUs, XPUs, TPUs of various kinds. Those are the near-term opportunities.
Okay. Okay, great. The NBMs, when do you think you'll start seeing some kind of significant revenue from those?
We have a lot of design wins. By the way, automotive is another opportunity for the NBM, but we have quite a few wins in a variety of other end applications. That too is a device we're now upgrading to 4G, and that will make it even more efficient and more capable and even more cost-effective. We're moving across all of these opportunities, from a product development perspective bringing about a next generation control technology that provides for more flexibility, more performance, lower cost, lower part count. We see these initiatives as bearing fruit in the order that I've outlined a little while ago.
It'd be similar to the front-end type stuff, maybe hit a $10 million run in a year from now or so?
I think the opportunity for NBMs is quite significant. It's another example of a product capability without equal. It's far smaller than anything else out there'll be many tries, including some failed tries at that.
Okay.
Not from us.
Jamie, you mentioned that if the tariff stuff goes away, you might see an acceleration of bookings. Does that mean we could possibly, and I know you're not going to commit to this, but we could possibly see numbers better than we talked about earlier if the tariffs go away?
Let's not speculate about what might happen if the tariffs go away in terms of impact on near-term bookings. I don't know what the lag would be, but I don't think there would be a large lag, and certainly this would be a very beneficial development in terms of our penetration of the Chinese market, which is a very substantial market.
Great. All right. Thank you very much, guys. Appreciate it.
Thank you. We'll be talking to you in a few months.